Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Cover photo: setting a modified fyke net in the Niagara River during 2011 brown bullhead sampling
efforts. U.S. Fish and Wildlife Service Photo.
Table of Contents
I. Executive Summary ..................................................................................................... 1
II. Background ................................................................................................................. 3
A. Delisting Criterion .................................................................................................... 3
B. Endpoint .................................................................................................................. 3
III. Impairment Status Resolution .................................................................................... 5
A. Strategy and Rationale ............................................................................................ 5
Development of Presque Isle Bay Methodology ...................................................... 5
Identification of a Least Impacted Control Site ......................................................... 6
Tumors in Bullhead as an Environmental Indicator .................................................. 7
B. Supporting Data and Assessment ........................................................................... 8
Finding of Impairment .............................................................................................. 8
Canadian Evaluation of Impairment ......................................................................... 9
U.S. Fish and Wildlife Service Study ........................................................................ 9
Statistical Analysis ................................................................................................. 13
C. Criteria, Principles and Guidance Application ....................................................... 17
D. Removal Statement ............................................................................................... 18
IV. BUI Removal Steps and Follow-up .......................................................................... 19
A. BUI Removal Steps ............................................................................................... 19
B. Post-Removal Responsibilities .............................................................................. 19
New York State Department of Environmental Conservation ................................. 20
United States Environmental Protection Agency .................................................... 20
V. References ............................................................................................................... 21
Appendix A – List of Remedial Advisory Committee Members
Appendix B – Public Meeting Notes
Appendix C – Responsiveness Summary
Appendix D – Statistical Analysis Report
List of Tables
Table 1 – Historical Liver Neoplasm Prevalence in Niagara River Brown Bullhead ........ 9
Table 2 – Sampling Sites .............................................................................................. 11
Table 3 – Liver Tumors by Site ..................................................................................... 11
Table 4 – Tumor Prevalence Estimates for Age 6 Fish ................................................. 16
List of Figures
Figure 1 – U.S. Niagara River Area of Concern .............................................................. 4
Figure 2 – Upper River Sampling Areas ........................................................................ 10
Figure 3 – Lower River Sampling Areas ........................................................................ 12
Figure 4 – Graphical representation of the hierarchical model ...................................... 14
Figure 5 – Example Probability Distribution ................................................................... 15
I. Executive Summary 1
I. Executive Summary
Based on the results of a U.S. Fish and Wildlife Service (USFWS) study of the
prevalence of liver tumors in Niagara River brown bullhead, the New York State
Department of Environmental Conservation (NYSDEC), with the support of the Niagara
River Remedial Advisory Committee, is recommending a change in the status of the
Fish Tumors or Other Deformities beneficial use impairment (BUI) from “impaired” to
“not impaired” for the U.S. side of the Niagara River Area of Concern (AOC). This report
explains the background and history of the Niagara River fish tumors BUI, and
discusses the strategy and rationale for the USFWS study along with its results.
In the Great Lakes Water Quality Agreement, the United States and Canada defined a
set of 14 BUIs that may exist in the Great Lakes AOCs as a result of chemical, physical
or biological disturbances to the ecosystem. Fish Tumors or Other Deformities is one of
the 14 potential BUIs. Ontario and New York State independently manage their
respective portions of the Niagara River AOC. On the U.S. side, the BUI was originally
given the status of “impaired” based on evidence from studies conducted in the 1980s
of higher than normal tumor rates in the river’s fish. On the Canadian side, the BUI was
originally given a status of “requiring further assessment.” Based on the results of a
Canadian study completed in 2010, also focused on the prevalence of liver tumors in
brown bullhead, the Canadian RAP Coordinating Committee changed the status of the
BUI to “not impaired” for the Canadian side.
Re-designation of a BUI’s status from impaired to not impaired may occur when
conditions within the AOC meet the approved delisting criteria for that BUI. In the case
of the Fish Tumors or Other Deformities BUI, the Niagara River Remedial Advisory
Committee approved a single delisting criterion. Based on the recommendations of
experts, the criterion requires that the prevalence of liver tumors in brown bullhead in
the AOC be statistically equivalent to or lower than the prevalence at a control site,
Long Point Inner Bay in Ontario. The selection of the control site and the methodology
of the USFWS study were based on the extensive efforts of a number of parties in
addressing this BUI for the Presque Isle Bay AOC in Pennsylvania.
The USFWS carried out its study specifically to evaluate whether Fish Tumors or Other
Deformities continues to be an impairment on the U.S. side of the Niagara River AOC.
The study began in 2011. USFWS conducted sampling over three years, collecting 50
fish each from upper, middle and lower zones of the AOC. Examination of livers at the
U.S. Geological Survey’s National Fish Health Research Laboratory showed that five of
the 150 fish collected for the study had liver tumors, for an overall uncorrected tumor
prevalence of 3.3 percent.
I. Executive Summary 2
Because the probability of an individual fish having a tumor depends on its age and
gender, an accurate comparison of tumor prevalence between the AOC and the control
site required the use of a statistical model. The model took into account the age and
gender of the sampled fish, as well as effects caused by sampling in multiple years and
multiple locations within the AOC. Statistical analysis using this model showed that the
tumor prevalence within the Niagara River AOC is statistically equivalent to the
prevalence at Long Point Inner Bay.
Based on this result, the delisting criterion for the Fish Tumors or Other Deformities BUI
has been met. Accordingly, NYSDEC and the Remedial Advisory Committee fully
support the re-designation of its status from “impaired” to “not impaired.”
II. Background 3
II. Background
The Niagara River is an AOC as defined in the Great Lakes Water Quality Agreement
(GLWQA) between the United States and Canada. The binational AOC encompasses
the entire River on both sides of the international border. Along New York’s coast, the
AOC extends from the mouth of Smoke Creek near the southern end of Buffalo Harbor
north to the mouth of the Niagara River at Lake Ontario (Figure 1). It also includes the
lower reaches of five tributaries: Smoke Creek, South Branch Smoke Creek,
Scajaquada Creek, Cayuga Creek and Gill Creek. Ontario and New York State
independently developed Remedial Action Plans (RAPs) for their respective portions of
the River.
In the GLWQA, the United States and Canada defined 14 potential BUIs for AOCs that
could result from changes in the chemical, physical or biological integrity of the Great
Lakes System. In order for the United States or Canada to delist an AOC, it must first
document the restoration of each of the BUIs.
One of the 14 potential impairments that is present in the Niagara River is Fish Tumors
or Other Deformities. This report outlines the available data addressing the status of this
BUI, and documents the Remedial Advisory Committee’s evaluation of the data and its
recommendation to re-designate the status of this BUI from “Impaired” to “Not Impaired”
(also referred to as removing the BUI).
A. Delisting Criterion
As approved by the Remedial Advisory Committee in December 2012, the delisting
criterion for the Fish Tumors or Other Deformities BUI is:
The prevalence of neoplastic liver tumors in brown bullhead is statistically
equivalent to or lower than the prevalence at Long Point Inner Bay (as
confirmed by histopathology).
B. Endpoint
The desired endpoint is to demonstrate that the prevalence of neoplastic liver tumors in
brown bullhead throughout the Niagara River AOC is statistically equivalent to the
prevalence at a least impacted control site. Long Point Inner Bay has been shown to be
a least impacted control site for Lake Erie (Rutter 2010).
III. Impairment Status Resolution 5
III. Impairment Status Resolution
A. Strategy and Rationale
The United States Environmental Protection Agency (USEPA) Delisting Guidance
document, Restoring United States Great Lakes Areas of Concern: Delisting Principles
and Guidelines, adopted by the United States Policy Committee (USPC 2001) states
the following:
“Re-designation of a BUI from impaired to unimpaired can occur if it can be
demonstrated that:
- Approved delisting criteria for that BUI have been met;
- The impairment is not solely of local geographic extent, but is typical of
upstream conditions OR conditions outside of the AOC boundaries on a
regional scale. Such re-designation would be contingent upon evidence
that sources within the AOC are controlled;
- The impairment is due to natural rather than human causes.”
The strategy in this case is to demonstrate that the delisting criterion has been met. Due
to the extensive efforts of a number of parties in developing a methodology to address
the Fish Tumors or Other Deformities BUI for the Presque Isle Bay AOC, NYSDEC and
USFWS elected to adopt the same methodology for the Niagara River AOC.
Development of Presque Isle Bay Methodology
International Joint Commission (IJC) delisting guidelines state that Fish Tumors or
Other Deformities may be deemed to be not impaired “when the incidence rates of fish
tumors or other deformities do not exceed rates at unimpacted control sites or when
survey data confirm the absence of neoplastic or pre-neoplastic liver lesions in
bullheads or suckers” (IJC 1991). However, these early guidelines did not address the
definitions of “unimpacted control sites” and “pre-neoplastic liver tumors,” or how to
determine background tumor rates.
The status of the fish tumor BUI for the Presque Isle Bay AOC was changed to “in
recovery” in 2002 based on studies showing declining rates of liver and skin tumors in
brown bullhead. This meant that all necessary remedial activities were thought to be
complete and the focus would turn to monitoring and determining when removal of the
BUI would be appropriate. As the Pennsylvania Department of Environmental Protection
III. Impairment Status Resolution 6
(DEP), Pennsylvania Sea Grant, the Public Advisory Committee, and other involved
parties considered the issue of when to remove the BUI, more questions arose that the
studies done in Presque Isle Bay or elsewhere in the Great Lakes could not answer
(Rafferty et al. 2009).
To begin answering these questions and to address past inconsistencies in the
evaluation of the fish tumor impairment across the Great Lakes, the Pennsylvania DEP,
U.S. Environmental Protection Agency Great Lakes National Program Office (GLNPO),
and Pennsylvania Sea Grant sponsored three conferences related to fish tumors in
Great Lakes AOCs during 2003 through 2006. Pennsylvania Sea Grant recruited a
panel of national experts in fish pathology and biological assessment to form a core
group for development of criteria and sampling methods. Subcommittees of the
conference participants developed a field manual on collecting and processing brown
bullhead (Rafferty and Grazio 2006) and a manual on diagnosing tumors in brown
bullhead (Blazer et al. 2007). The purpose of the latter manual was to describe
neoplastic and non-neoplastic proliferative lesions of the liver and skin of the brown
bullhead and suggest terminology that could be consistently used at AOCs throughout
the Great Lakes and in other areas.
Experts also published recommendations on sampling frequency and distribution within
an AOC for studies to determine tumor prevalence (Blazer et al. 2009). They noted that
localized differences in tumor prevalence were found in the Presque Isle Bay and
Cuyahoga River AOCs, and suggested that areas of high sediment, water or dietary
contaminant loads may be important. They also noted annual variation in liver
neoplasms at Presque Isle Bay and at some potential reference sites. The reasons for
the variation were not identified, but could include sample size, the particular location
within an AOC or reference site at which fish were collected, age of fish collected, and
environmental factors (both recent and early life stage exposures). Based on these
observations, they recommended that sampling should be conducted during more than
one year and at multiple sites within a large AOC to obtain an accurate estimate of
tumor prevalence. These recommendations, together with the field and diagnostic
manuals, provide the rationale for the design of the USFWS study of tumor prevalence
in the Niagara River AOC.
Identification of a Least Impacted Control Site
In order to identify a least-impacted control site, the Pennsylvania DEP sampled brown
bullhead from a number of candidate sites across Lake Erie. In addition to having a
resident bullhead population, candidate sites had to be outside any AOC and could not
have point-source discharges of pollutants or known sediment contamination. Based on
III. Impairment Status Resolution 7
these criteria, the DEP collected samples in 2004, 2005, and 2007 from Dunkirk Harbor
(New York), Long Point Inner Bay (Ontario), Old Woman's Creek (Ohio), and Sandusky
Bay (Ohio) (PADEP 2012).
The evaluation of the candidate sites considered prevalence of both liver and skin
tumors. Dunkirk Harbor had the lowest estimated liver tumor prevalence (0.0%), while
Long Point Inner Bay (LPIB) had the lowest estimated skin tumor prevalence (6.4%).
However, the estimate for the Dunkirk Harbor liver tumor prevalence had a high level of
uncertainty, with the 95% confidence interval ranging from 0.0% to 56.0%. Dunkirk
Harbor also had the highest estimated prevalence of skin tumors (22.5%). LPIB had the
second lowest estimated liver tumor prevalence (1.2%) and the 95% confidence interval
was much narrower (0.0% to 14.9%), indicating less uncertainty in the estimate.
Therefore, LPIB was selected as the least-impacted Lake Erie control site (Rutter 2010).
Tumors in Bullhead as an Environmental Indicator
Brown bullhead catfish are commonly accepted as good indicators of local
environmental conditions because they are a bottom-dwelling fish and are considered to
have a limited home range. Of tumors that occur in brown bullhead, the types that
experts have most commonly associated with exposure to contaminants are liver
tumors and, to a lesser extent, skin tumors (Rafferty et al. 2009).
Numerous studies have provided evidence for a link between exposure to chemicals,
most often polycyclic aromatic hydrocarbons (PAHs), in the sediment of lakes and rivers
and an increased prevalence of liver tumors in brown bullhead (Rafferty et al. 2009).
However, these studies cannot prove a definite cause and effect relationship. In fact,
bullheads with tumors have been found in both contaminated and uncontaminated
waterbodies throughout the northeastern United States (PADEP 2012). However,
Rafferty et al. (2009) concluded that the weight of evidence does suggest an
association between exposure to PAHs and liver tumors in brown bullhead. They added
that PAHs may not be the only class of chemicals responsible for such tumors, though
more research is needed regarding other chemicals.
Rafferty et al. (2009) also concluded that the weight of evidence for an association
between exposure to PAHs and skin tumors is “much more problematic.” The
Pennsylvania DEP later conducted its own investigation into the cause of skin tumors
while in the process of determining whether to delist the Presque Isle Bay AOC. Their
goal was to evaluate whether skin tumors are an appropriate indicator of environmental
contamination. Their efforts included an investigation of pathogens as potential natural
causes of the tumors, an evaluation of the possible role of genetics, an experimental
III. Impairment Status Resolution 8
investigation for any relationship between exposure to Bay sediment and the
development of tumors, and an extensive literature review. These efforts were unable to
isolate any factors responsible for the development of skin tumors. This result, coupled
with the fact that their studies at multiple locations had not shown a correlation between
high liver tumor prevalence and high skin tumor prevalence, led to the conclusion that
skin tumors are not a reliable indicator of environmental contamination (PADEP 2012).
In summary, while some debate continues regarding the association between liver
tumors and environmental contamination, liver tumors have become widely accepted as
an indicator for the purpose of evaluating the Fish Tumors BUI. Skin tumors, however,
are not generally accepted as an appropriate indicator. As previously noted, the Niagara
River Remedial Advisory Committee has approved a delisting criterion that considers
only liver tumors.
B. Supporting Data and Assessment
Finding of Impairment
Fish Tumors or Other Deformities was found to be an impairment on the U.S. side of the
Niagara River AOC based on limited evidence. The Niagara River Remedial Action Plan
(RAP) states that this finding is based on the results of two studies of tumor incidence in
fish from the upper river (NYSDEC 1994). A freshwater drum skin tumor study indicated
a skin tumor rate of 11.7% in fish collected from a location just downstream of Black
Rock Canal, higher than the Lake Erie-wide incidence level of 1.5% (Black 1983).
The second study compared the incidence of several types of abnormalities, including
neoplastic and pre-neoplastic liver tumors, in brown bullhead from the 102nd Street
embayment and a reference site on Black Creek, Ontario, a tributary to the Niagara
River (Hickey et al. 1990). Of 101 fish collected from the 102nd Street embayment,
three showed pre-neoplastic liver tumors and two had neoplastic liver tumors (Table 1).
(Contaminated sediment was removed from the embayment in 1996.) The RAP states
that these five tumors indicate “elevated conditions above the suggested zero rate of
this indicator of impairment.” The RAP also notes that due to some limitations of the
Hickey et al. (1990) study, further research would be necessary to comprehensively
define the extent of this impairment in the Niagara River.
The “suggested zero rate” for liver tumors mentioned in the RAP is no longer consistent
with scientific consensus. The background neoplastic liver tumor prevalence in brown
bullhead is now considered to range from two to five percent in fish age 3 and older
III. Impairment Status Resolution 9
(Baumann 2010). Therefore, by current standards the Hickey et al. (1990) study does
not provide strong evidence of a fish tumor beneficial use impairment.
Table 1 – Historical Liver Neoplasm Prevalence in Niagara River Brown Bullhead
Year of
Sampling
Sample
Size Location
Percentage with
Liver Tumors Source
1987 101 102nd Street embayment 2.0 Hickey et al. 1990
1998 40 102nd Street embayment 7.5 Blazer et al. 2009
2004 43 Queenston 0.0 Baumann 2010
2004 & 2008 101 Black Creek (Ontario) 3.0 Baumann 2010
Canadian Evaluation of Impairment
In the Canadian RAP, the Fish Tumors or Other Deformities BUI was originally given a
status of “Requiring Further Assessment.” In 2009, Environment Canada commissioned
a study by Dr. Paul Baumann to evaluate the status of this BUI for all of the Canadian
AOCs. The evaluation was based on the prevalence of liver tumors in brown bullhead.
For the Niagara River AOC, a total sample of 101 bullhead had been collected during
2004 and 2008 near Black Creek (Figure 2). Three of these fish had liver tumors (Table
1). This prevalence was not significantly different from the background prevalence that
Dr. Baumann had calculated based on data from a number of reference sites (Baumann
2010). Therefore, the Canadian RAP Coordinating Committee changed the status of the
Fish Tumors BUI to “Not Impaired” for the Canadian side.
U.S. Fish and Wildlife Service Study
The USFWS began a study of tumor prevalence in U.S. Niagara River AOC brown
bullhead in June 2011. For purposes of sampling, the river was divided into upper,
middle and lower zones with a planned sample size of 50 fish from each zone (Table 2).
The upper zone was separated into two sites, the developed shoreline (Buffalo Outer
Harbor and Black Rock Canal) and undeveloped shoreline (Strawberry Island and East
River Marsh). The middle zone was also divided into two sites, an area near the mouth
of Gun Creek at the northeast end of Grand Island and an area around Cayuga Island
that is in the vicinity of the Love Canal and 102nd Street Landfill hazardous waste sites
(Figure 2).
III. Impairment Status Resolution 11
Table 2 – Sampling Sites
Location Sampling Date Sample Size
Middle Zone
Near mouth of Gun Creek, Grand Island 6/14/2011 26
Cayuga Island area 6/21/2011 24
Upper Zone
Developed areas 6/12/2012 25
Undeveloped areas 6/13/2012 25
Lower Zone 7/23/2013 50
The lower zone included the portion of the river between Lewiston and Fort Niagara.
Shoreline along this approximately 6.7 mile reach is not heavily developed. Although
samples were collected in four distinct areas of the reach (Figure 3), the fish were not
kept separate for analysis because the habitat and the degree of shoreline development
were not remarkably different among the four areas.
U.S. Geological Survey (USGS) fishery biologist Dr. Vicki Blazer was responsible for the
examination of the fish for the presence of tumors. Necropsy teams visited the AOC to
process the brown bullhead samples. The histopathology work took place at the
USGS’s National Fish Health Research Laboratory in West Virginia.
Five of the 150 fish collected for the study were found to have neoplastic liver tumors
(Blazer et al. 2014). Three of these fish were collected in the middle zone and two in the
upper zone. The incidence of liver tumors ranged from 0% for the 2013 lower zone
sample to 7.7% for the 2011 Gun Creek sample (Table 3). The overall uncorrected
incidence was 3.3%. The mean age of the samples by location ranged from 4.2 to 6.8
years.
Table 3 – Liver Tumors by Site
Location Sample Size Number with
Liver Tumors
Percentage with
Liver Tumors
Male Female Male Female Male Female Combined
Middle Zone
Gun Creek area 18 8 2 0 11.1% 0 7.7%
Cayuga Island area 14 10 0 1 0 10% 4.2%
Upper Zone
Developed areas 13 12 0 1 0 8.3% 4.0%
Undeveloped areas 16 9 0 1 0 11.1% 4.0%
Lower Zone 36 14 0 0 0 0 0
Total 97 53 2 3 2.1% 5.7% 3.3%
III. Impairment Status Resolution 13
Statistical Analysis
The statistical methodology used to compare the Niagara River AOC tumor prevalence
to that of the control site (LPIB) follows the methodology used for the Presque Isle Bay
analysis (Rutter 2010). The full statistical analysis report can be found in Appendix D.
The methodology is complex, but a simple comparison of overall tumor prevalence
percentage between the AOC and the control site is not appropriate for two reasons:
It would involve an assumption that every fish collected for analysis has an equal
probability of having a tumor, which is not the case. The probability will vary with
one or more characteristics such as the age, length, weight and gender of
individual fish.
It would also involve an assumption of having a “simple random sample” of fish
from each site. Almost all statistical techniques assume a simple random sample.
However, because sampling occurred in multiple locations within the AOC and in
multiple years, this assumption is violated. As a result, true confidence intervals
for estimates of tumor prevalence become larger than one would calculate under
the simple random sample assumption. (Note that for the control site, fish were
sampled in the same location, but over multiple years.)
To address the first concern, creation of a model that accounts for the impact of the
demographics (age, length, gender and weight of fish) on the probability of a fish having
a tumor is necessary. In this case, a logistic regression model is appropriate because
the response variable in the analysis has only two possible outcomes, the presence or
absence of a tumor. Logistic regression can incorporate the effects of any number of
demographic characteristics (called predictor variables) on the probability of a fish
having a tumor.
Addressing the second concern requires the use of a hierarchical modelling approach to
represent the sampling design. The hierarchical approach accounts for the (slight)
correlation among fish sampled in the same location and year. The overall effect of a
predictor variable (e.g. age) on the probability of a fish having a tumor for the Niagara
River is modelled as a combination of the effect of that variable specific to each
sampling year (2011 – 2013). For 2011 and 2012, the year-specific effects are modelled
as a combination of the effects specific to each sampling location within those years.
Therefore, the hierarchy has an overall Niagara River effect at the top, which depends
on year-specific effects at the next level, some of which depend on location specific
effects at the bottom level (Figure 4).
III. Impairment Status Resolution 14
Figure 4 – Graphical Representation of the Hierarchical Model
Statistical modelling involves using observed data to estimate the model’s unknown
parameters, which are measures of the effects of the predictor variables (or
demographic characteristics in this case). This is known as “fitting” the model. In other
words, the goal is to find the set of parameter values that result in the closest match
between the model’s predictions and the observed data.
For technical reasons, the model parameters in this analysis were estimated using a
Bayesian framework. For each of the Niagara River areas sampled, no tumors were
found in fish of at least one gender (Table 3). The mathematics of the logistic regression
model are not workable for cases where no tumors are observed in a sample. While at
least one non-Bayesian method can address this problem, no available statistical
software can apply that method together with a traditional “mixed model” to account for
the sampling design. The Bayesian approach easily overcomes the problem with logistic
regression.
When comparing tumor prevalence between two sites, ensuring that the amount of data
is sufficient to detect meaningful differences is important; this is the “power” of the
statistical procedure. Power is a measure of the ability of a method to find a statistical
difference when a difference actually exists (i.e. to avoid a “false negative” result).
Greater statistical power is associated with larger sample sizes. Ensuring sufficient
power is important in this analysis so that a finding of no statistical difference in tumor
prevalence between the AOC and control site will be due to an actual small difference in
Niagara River Effect
Niagara River 2011 Effect
Niagara River 2012 Effect
Niagara River 2013 Effect
Niagara River 2012
Developed Areas Effect
Niagara River 2012
Undeveloped Areas Effect
Niagara River 2011
Gun Creek Area Effect
Niagara River 2011
Cayuga Island Area Effect
III. Impairment Status Resolution 15
tumor prevalence and not due simply to small sample sizes. No available statistical
software can perform a power analysis for a traditional mixed-model logistic regression.
Use of the Bayesian framework essentially transforms the power measurement into the
width of a confidence interval, with a smaller interval width indicating greater statistical
power. So specifying a maximum acceptable width of a confidence interval is equivalent
to requiring sufficient power.
The first phase of carrying out the analysis was to determine which combination of
predictor variables best described the observed tumor prevalence. Hierarchical logistic
regression models using all possible combinations of the four predictor variables (age,
length, weight and gender) were tested using the Bayesian framework. The best model
of those examined included age and gender as predictor variables.
The tumor prevalence for the control site and the Niagara River were then estimated. All
the sampling sites for the Niagara River were combined over all the years to arrive at a
single estimate of the AOC tumor prevalence. Similarly, a single tumor prevalence
estimate was generated for the control site, combining over years. Because the
probability of having a tumor varies with the age of the fish, it was necessary to choose
an age at which to calculate the estimates. The average age of bullhead in the data set
was 6.4 years, so all tumor prevalence estimates were calculated for age 6 bullhead.
Gender was the other important predictor variable, so estimates for both male and
female bullhead were calculated.
While a traditional statistical method would generate a point estimate for the desired
result, Bayesian analysis generates a probability distribution over the range of possible
result values. For this analysis, each probability distribution was summarized by finding
its median and the 95% highest probability density interval (HPDI), which is the smallest
interval within the range of possible result values that is associated with a total
probability of 95%. Figure 5 provides
an illustration for an arbitrary
probability distribution. The horizontal
axis represents possible result values
and the vertical axis represents
probability. The median is the result
value for which the total probability
associated with all lower values and
the total probability associated with
all higher values are both equal to
50%. The 95% HPDI, which is
shaded, is denoted as (a, b).
Figure 5 – Example Probability Distribution
III. Impairment Status Resolution 16
In this analysis, the median of the probability distribution is used as a point estimate of
tumor prevalence while the 95% HPDI is a Bayesian analogue of a 95% confidence
interval. The resulting point estimates of Niagara River AOC tumor prevalence (Table 4)
for both female fish (2.0%) and male fish (1.9%) compare favorably with the generally
accepted background prevalence mentioned earlier of two to five percent for fish age 3
and older.
Table 4 – Tumor Prevalence Estimates for Age 6 Fish
Location and Gender 95% HPDI Median
Niagara River: Female (0.0%, 25.5%) 2.0%
Niagara River: Male (0.0%, 35.2%) 1.9%
Long Point: Female (0.0%, 4.4%) 0.4%
Long Point: Male (0.0%, 5.1%) 0.3%
The final step of the analysis was to conduct a statistical test for equivalency between
the tumor prevalence at the AOC and at LPIB using a Two One-Sided Tests (TOSTs)
procedure, which is accepted by the U.S. Food and Drug Administration for evaluating
the comparability between two groups. Hypothesis testing commonly involves a null
hypothesis that no difference exists in some quantity measured for two groups and an
alternative hypothesis that a difference does exist. However, for the TOST procedure,
the null hypothesis is that the measurements for two groups are not equivalent or,
stated another way, that the difference in the measurements exceeds some pre-defined
tolerance level. For this analysis, the null hypothesis can be stated as
H0: PREVLPIB – PREVAOC ≤ −𝜃 or PREVLPIB – PREVAOC ≥ 𝜃
where PREVLPIB and PREVAOC are the tumor prevalence at LPIB and the AOC
respectively and 𝜃 is the tolerance level. The alternative hypothesis is
HA: −𝜃 < PREVLPIB – PREVAOC < 𝜃
The TOST procedure is performed as two separate one-sided tests having null and
alternative hypotheses:
H0: PREVLPIB – PREVAOC ≤ −𝜃
HA: PREVLPIB – PREVAOC > −𝜃
H0: PREVLPIB – PREVAOC ≥ 𝜃
HA: PREVLPIB – PREVAOC < 𝜃
III. Impairment Status Resolution 17
The p-value is the larger of those obtained for the two one sided tests. If the p-value is
less than an acceptable significance value (chosen to be 0.05) then sufficient evidence
exists to suggest the tumor prevalences are equivalent at the 𝜃 tolerance level.
An equivalent approach to the TOST procedure is to construct a 90% confidence
interval for the difference in tumor prevalence between the AOC and control site. (Under
the Bayesian framework, the analogous 90% HPDI for the difference in tumor
prevalences was determined.) If this interval is entirely between the tolerance limits −𝜃
and 𝜃, then the prevalences can be considered to be equivalent. If the interval is too
large or does not contain zero, then the prevalences are statistically significantly
different. The tolerance is determined essentially by applying the TOST procedure to
compare the AOC to itself.
The Niagara River AOC tolerance intervals were determined to be ±24% for female fish
and ±34% for male fish. Comparing Niagara River to LPIB, the calculated 90% TOST
HPDI interval for tumor prevalence in females is (-5.4%, 18.8%) and in males is (-5.5%,
25.5%). Both of these results are within the corresponding tolerance intervals, indicating
statistical equivalency between the two sites.
C. Criteria, Principles and Guidance Application
The intent of the RAP process is to assess the status of each BUI and, if existence of an
impairment is indicated, to remedy the source of the impairment and subsequently
demonstrate that the beneficial use has been restored. In this case, initial evidence for
an impairment was limited. Also, contaminant concentrations in the Niagara River have
fallen significantly since the time of the two studies that indicated an impairment.
However, the relationship between contaminants found in the Niagara River and brown
bullhead tumors is uncertain. Therefore, conducting a new study to determine whether
tumor prevalence is elevated in comparison to a control site was appropriate.
The delisting criterion for the Fish Tumors or Other Deformities impairment is met
because the USFWS fish tumor study, conducted according to the guidance prepared
by experts in the fields of fish pathology and biological assessment, has shown
statistical equivalency between the tumor prevalence within the Niagara River AOC and
the LPIB control site.
III. Impairment Status Resolution 18
D. Removal Statement
Based on the demonstrated equivalency between the prevalence of liver tumors in the
Niagara River and the prevalence in a least impacted control site, the endpoint for the
Fish Tumors or Other Deformities BUI has been met. Accordingly, NYSDEC and the
Remedial Advisory Committee fully support the re-designation of its status from
“impaired” to “not impaired.”
IV. BUI Removal Steps and Follow-up 19
IV. BUI Removal Steps and Follow-up
A. BUI Removal Steps
Completed Date Action
1. ✔ 12/2008 Delisting criteria completed and finalized with USEPA
2. ✔ 2/2010 USFWS proposed a multi-year study to measure the prevalence of tumors in brown bullhead in the U.S. portion of the Niagara River AOC
3. ✔ 1/2012 Original impaired condition reviewed to identify causes and sources
4. ✔ 12/2012
Remedial Advisory Committee endorsed a change to the BUI delisting criterion for consistency with the Pennsylvania Department of Environmental Protection’s approach for the Presque Isle Bay AOC
5. ✔ 5/2014 USGS issued a report on the Niagara River brown bullhead analysis
6. ✔ 11/2014 Dr. Rutter completed a preliminary statistical analysis showing that tumor prevalence in the AOC is equivalent to that in the control site
7. ✔ 7/2015 Discussion of removal by the Remedial Advisory Committee
8. ✔ 7/2015 NYSDEC collaborated with USEPA to revise the draft BUI removal proposal document
9. ✔ 8/2015 - 9/2015
NYSDEC conducted outreach and held a public meeting to solicit comments on the draft removal proposal (30-day public comment period)
10. ✔ 10/2015 NYSDEC assembles comments, prepares responsiveness summary and completes final modifications to the BUI removal document
11. 11/2015 Coordinate the formal transmittal of the BUI removal with USEPA GLNPO. Communicate result with IJC
12. 11/2015 Communicate results to local RAP Coordination for appropriate recognition and follow-up
B. Post-Removal Responsibilities
Following removal of the Fish Tumors or Other Deformities BUI, NYSDEC and USEPA
will continue ongoing environmental programs that will ensure that the BUI’s status
remains “not impaired.” These programs include hazardous waste site remediation,
contaminated sediment remediation, monitoring of permitted wastewater discharges,
IV. BUI Removal Steps and Follow-up 20
water quality monitoring, and coordination of the Niagara River Remedial Advisory
Committee.
Regular monitoring of water quality will continue under the Niagara River Toxics
Management Plan, operated by a partnership of USEPA, NYSDEC, Environment
Canada and Ontario Ministry of the Environment and Climate Change. This monitoring
includes biweekly sampling at both Fort Erie and Niagara-on-the-Lake to check
concentrations of a range of potential contaminants.
New York State Department of Environmental Conservation
Under the State Superfund Program, NYSDEC will continue working to complete
the remediation of hazardous waste sites. Remediation is complete at 37 of 44
sites originally found to have the potential to impact the Niagara River. Remedial
activities are under way at the seven remaining sites.
Through the State Pollutant Discharge Elimination System, NYSDEC will
continue to regulate point source discharges of industrial and municipal
wastewater and storm water in accordance with the federal Clean Water Act.
NYSDEC will continue to monitor water quality in the AOC through its statewide
Rotating Integrated Basin Studies (RIBS) ambient water quality monitoring
program. NYSDEC collects biological samples every five years at several
Niagara River tributary locations, and conducts routine monitoring on the Niagara
River near Fort Niagara 5 – 6 times per year in spring, summer, and fall. The
samples are analyzed for a wide range of potential contaminants and toxicity
bioassays using C. dubia are conducted every five years.
NYSDEC will continue to provide a coordinator for the Niagara River AOC whose
responsibilities include coordination of the Remedial Advisory Committee.
United States Environmental Protection Agency
Under the Great Lakes Legacy Act Program, USEPA will continue to lead efforts
to characterize sediment within the AOC, and to conduct remediation where
necessary.
USEPA will also continue to provide funding for technical assistance on AOC
projects to the extent that resources are available.
V. References 21
V. References
Baumann, Paul C. 2010. Data Analysis and Fish Tumor BUI Assessment for the Lower
Great Lakes and Interconnecting Waterways. Submitted to Environment Canada,
March 31, 2010.
Black, J.J. 1983. Field and Laboratory Studies of Experimental Carcinogenesis in the
Niagara River. Journal of Great Lakes Research. 9(2): 326-334.
Blazer, V.S., J.W. Fournie, J.C. Wolf, and M.J. Wolf. 2007. Manual for the Microscopic
Diagnosis of Proliferative Liver and Skin Lesions in the Brown Bullhead (Ameiurus
Nebulosus). Pennsylvania Sea Grant. 34pp.
Blazer, V., S. Rafferty, C. Baumman, S. Smith, and E. Obert. 2009. Assessment of the
“fish tumors or other deformities” beneficial use impairment in brown bullhead
(Ameiurus nebulosus): II. Liver neoplasia. Journal of Great Lakes Research. 35(4): 527-
537.
Blazer, V.S., Mazik, P.M., Iwanowicz, L.R., Braham, R.P., Hahn, C.M., Walsh H.L., and
Sperry, A.J., 2014, Assessment of the fish tumor beneficial use impairment in brown
bullhead (Ameiurus nebulosus) at selected Great Lakes Areas of Concern: U.S.
Geological Survey Open-File Report 2014-1105, 17 p.,
http://dx.doi.org/10.3133/ofr20141105.
Hickey J.T., Bennett R.O., and Merckel C. 1990. Biological Indicators of Environmental
Contaminants in the Niagara River: Histological Evaluation of Tissues from Brown
Bullheads at the Love Canal-102nd Street Dump Site Compared to the Black Creek
Reference Site. U.S. Fish and Wildlife Service, Cortland, New York.
International Joint Commission. 1991. Proposed listing/delisting guidelines for Great
Lakes Areas of Concern. Focus on International Joint Commission Activities. Volume
14, Issue1, Insert.
New York State Department of Environmental Conservation (NYSDEC). 1994. Niagara
River Remedial Action Plan. September 1994.
Pennsylvania Department of Environmental Protection (PADEP). 2012. Presque Isle
Bay Area of Concern Final Stage 3 Remedial Action Plan: Delisting. December 2012.
Rafferty, S.D. and J.L. Grazio. 2006. Field Manual for Assessing Internal and External
Anomalies in Brown Bullhead (Ameiurus nebulosus). Pennsylvania Sea Grant, 29 pp.
V. References 22
Rafferty S., V. Blazer, A. Pinkney, J. Grazio, E. Obert, and L. Boughton. 2009. A
historical perspective on the “fish tumors or other deformities” beneficial use impairment
at Great Lakes Areas of Concern. Journal of Great Lakes Research. 35(4): 496-506.
Rutter, MA. 2010. A statistical approach for establishing tumor incidence delisting
criteria in areas of concern: a case study. Journal of Great Lakes Research. 36(4): 646-
55.
USFWS. 2011. Quality Assurance Project Plan: Assessing the Status of the “Fish
Tumors and Other Deformities” Beneficial Use Impairment within the Niagara River
Area of Concern. June 2011.
USPC. 2001. Restoring United States Areas of Concern: Delisting Principles and
Guidelines. Adopted by the US Policy Committee on December 6, 2001
Appendix A – List of Remedial Advisory Committee Members A-1
Appendix A – List of Remedial Advisory Committee Members Sarah Delavan
Assistant Professor
State University of New York at Buffalo
230 Jarvis Hall
Buffalo, New York 14260-4400
(716) 645-1810
Bryan Hinterberger
US Army Corps of Engineers
Buffalo District
1776 Niagara Street
Buffalo, New York 14207
(716) 879-4409
James Devald
Director of Environmental Health
Niagara County Department of Health
5467 Upper Mountain Road, Suite 100
Lockport, New York 14094-1894
(716) 439-7444
Kerrie Gallo
Director of Ecological Programs
Buffalo Niagara Riverkeeper
721 Main Street
Buffalo, New York 14203
(716) 852-7483 x30
Helen Domske
Extension Specialist
New York Sea Grant/GLP
State University of New York at Buffalo
204 Jarvis Hall
Buffalo, New York 14260
(716) 645-3610
Alexander Karatayev
Director, Great Lakes Center
Science Building 261
Buffalo State College
1300 Elmwood Avenue
Buffalo, New York 14222
(716) 878-5423
Dimitry Gorsky
U.S. Fish & Wildlife Service
Lower Great Lakes Fish and Wildlife
Conservation Office
1101 Casey Road
Basom, New York 14013
(585) 948-5445 ext. 2229
Thomas Mayer
Superintendent of Manufacturing
Engineering
GM Powertrain Tonawanda Engine Plant
2995 River Road
Buffalo, New York 14207-1099
(716) 879-5799
Thomas Hersey
Deputy Commissioner
Division of Environmental Compliance
Department of Environment and Planning
Erie County
95 Franklin Street, Room 1085
Buffalo, New York 14202
(716) 858-7674
Tony Scime
Erie County Federation of Sportsmen
Clubs
242 Southwood Drive
Buffalo, New York 14223
(716) 877-1474
Appendix A – List of Remedial Advisory Committee Members A-2
Dennis Sutton
Buffalo Urban Renewal Agency
920 City Hall
65 Niagara Square
Buffalo, New York 14202-3376
(716) 851-6587
Dr. Brenda Young
Associate Professor
Daemen College
Department of Natural Sciences
4380 Main Street
Amherst, New York 14226
(716) 839-8366
Vacant
Buffalo Ornithological Society
Appendix B – Public Meeting Notes B-1
Appendix B – Public Meeting Notes
Meeting Summary
Niagara River Area of Concern
Fish Tumors or Other Deformities Beneficial Use Impairment Removal Proposal
Public Meeting
Tuesday, August 25, 2015
6:00 – 8:00 p.m.
Grand Island Memorial Library, Grand Island, New York
Participants:
NYSDEC – Shannon Dougherty, Don Einhouse, Mark Filipski, Jim Lehnen,
Paul McKeown, Don Zelazny
Public – Ron Rezabek (Niagara River Station Fishing Club)
Introduction:
Mr. Zelazny, NYSDEC Great Lakes Programs Coordinator, introduced himself as well
as Mr. Filipski, NYSDEC Niagara River Area of Concern Coordinator, and Mr. Einhouse,
Senior Great Lakes Fisheries Biologist for Lake Erie. He mentioned that this is
NYSDEC’s first public meeting on a beneficial use impairment removal proposal for the
Niagara River Area of Concern.
Presentation:
Mr. Filipski gave a presentation that included a brief overview of the Great Lakes Areas
of Concern program and the Niagara River Area of Concern in particular, followed by
specific information regarding the proposal to remove the fish tumors beneficial use
impairment (BUI). A summary of the presentation follows.
Areas of Concern and the Niagara River
In 1987, the United States and Canada agreed to give the most polluted areas of the
Great Lakes priority attention for restoration. They designated a total of 43 such areas
as “Areas of Concern” because they had serious pollution problems to a greater degree
than in the rest of the Great Lakes. The problems in these areas seriously restricted
Appendix B – Public Meeting Notes B-2
beneficial uses of the water. Twenty-six Areas of Concern were located entirely within
the United States, 12 within Canada, and five, including the Niagara River, were
binational waterways.
In the Great Lakes Water Quality Agreement, the United States and Canada defined 14
potential beneficial use impairments for Areas of Concern that could result from
changes in the chemical, physical or biological integrity of Great Lakes waters. The U.S.
side of the Niagara River AOC has 7 of the 14 possible BUIs. With the exception of
habitat loss, all of the Niagara River BUIs are caused at least in part by pollutants. The
habitat loss BUI is the result of physical changes to the river that accompanied
urbanization, and water level fluctuations associated with power production.
The Area of Concern includes the entire River. On the U.S. side, it extends from Smoke
Creek near the southern part of the Buffalo Harbor north to the mouth of the River at
Lake Ontario. It does not extend inland to include the watershed. Ontario and New York
developed separate Remedial Action Plans for the Canadian and U.S. sides of the
River. New York published its plan in 1994. The goal of the remedial action plan is to
restore all of the beneficial use impairments.
Sources of contaminants to the Niagara River include contaminated sediments,
hazardous waste sites, groundwater, wastewater discharges, sewer overflows, other
nonpoint sources and inflow from Lake Erie. Progress on addressing these sources
includes:
- A total of 44 U.S. hazardous waste sites were found to have the potential to
release contaminants to the River. Today, measures to address the
contamination are complete for 37 of the 44 sites. Actions are under way at all 7
remaining sites.
- Projects addressing sediment contamination have been completed at 14
locations, with a few of those locations having multiple projects. These projects
were generally components of a hazardous waste site cleanup and in total they
removed over 300,000 cubic yards.
- Upgrades to WWTPs have reduced the amount of contaminants entering the
river, especially from the City of Niagara Falls.
- Within the Niagara River watershed, the cities of Buffalo, Lockport and Niagara
Falls are implementing long-term control plans to reduce the frequency, duration,
and intensity of combined sewer overflows. Implementation of these plans is
Appendix B – Public Meeting Notes B-3
quite expensive and is therefore spread over a number of years, but the plans
are enforceable so they do guarantee reductions in this source of contaminants.
Fish Tumors or Other Deformities Impairment
A group of experts has been studying tumors in brown bullhead catfish for a number of
years. They have concluded that the weight of evidence suggests an association
between exposure to polycyclic aromatic hydrocarbons (PAHs) and liver tumors in
brown bullhead. The evidence for an association between exposure to PAHs and skin
tumors is much weaker. Liver tumors have become widely accepted as an indicator for
the purpose of evaluating the Fish Tumors BUI. Skin tumors, however, are not generally
accepted as an appropriate indicator. Therefore, although skin tumors played a role in
the original finding of a fish tumor impairment for the Niagara River, only liver tumors
are considered in the current evaluation of the impairment.
The original finding of an impairment on the U.S. side of the Niagara River was based
on limited evidence. The Niagara River Remedial Action Plan (RAP) states that this
finding is based on the results of two studies of tumor incidence in fish from the upper
river.
- A freshwater drum skin tumor study indicated a skin tumor rate of 11.7% in fish
collected from a location just downstream of Black Rock Canal, higher than the
Lake Erie-wide incidence level of 1.5%.
- The second study focused on brown bullhead from the 102nd Street embayment.
Of 101 fish collected, three showed pre-neoplastic liver tumors and two had
neoplastic liver tumors. The RAP states that these five tumors indicate “elevated
conditions above the suggested zero rate of this indicator of impairment.”
The “suggested zero rate” for liver tumors mentioned in the RAP is no longer consistent
with scientific consensus. The background liver tumor prevalence in brown bullhead is
now considered to range from two to five percent in fish age 3 and older.
In the Canadian RAP, the Fish Tumors BUI was originally given a status of “Requiring
Further Assessment.” In 2009, Environment Canada commissioned a study to evaluate
the status of this BUI for all of the Canadian AOCs. The evaluation was based on the
prevalence of liver tumors in brown bullhead. For the Niagara River AOC, a total sample
of 101 bullhead had been collected during 2004 and 2008 near Black Creek. Three of
these fish had liver tumors. This prevalence was not significantly different from the
background prevalence calculated based on data from a number of reference sites.
Appendix B – Public Meeting Notes B-4
Therefore, the Canadian RAP Coordinating Committee changed the status of the Fish
Tumors BUI to “Not Impaired” for the Canadian side.
The International Joint Commission published delisting guidelines in 1991 that stated
that Fish Tumors or Other Deformities may be deemed to be not impaired “when the
incidence rates of fish tumors or other deformities do not exceed rates at unimpacted
control sites or when survey data confirm the absence of neoplastic or pre-neoplastic
liver lesions in bullheads or suckers.” However, these early guidelines did not address
the definitions of “unimpacted control sites” and “pre-neoplastic liver tumors,” or how to
determine background tumor rates.
The status of the fish tumor BUI for the Presque Isle Bay AOC in Pennsylvania was
changed to “in recovery” in 2002 based on studies showing declining rates of liver and
skin tumors in brown bullhead. As involved parties considered the issue of when to
remove the BUI, more questions arose that the studies done in Presque Isle Bay or
elsewhere in the Great Lakes could not answer. To begin answering these questions
and to address past inconsistencies in the evaluation of the fish tumor impairment
across the Great Lakes, the Pennsylvania DEP, U.S. Environmental Protection Agency
Great Lakes National Program Office (GLNPO), and Pennsylvania Sea Grant
sponsored three conferences related to fish tumors in Great Lakes AOCs during 2003
through 2006. Pennsylvania Sea Grant recruited a panel of national experts in fish
pathology and biological assessment to form a core group for development of criteria
and sampling methods. Subcommittees of the conference participants developed a field
manual on collecting and processing brown bullhead and a manual on diagnosing
tumors in brown bullhead.
Experts also published recommendations on sampling frequency and distribution within
an AOC for studies to determine tumor prevalence. They recommended that sampling
should be conducted during more than one year and at multiple sites within a large AOC
to obtain an accurate estimate of tumor prevalence. These recommendations, together
with the field and diagnostic manuals, provided the rationale for the design of the
USFWS study of tumor prevalence in the Niagara River AOC.
In order to identify a least-impacted control site, the Pennsylvania DEP sampled brown
bullhead from a number of candidate sites across Lake Erie. The evaluation of the
candidate sites considered prevalence of both liver and skin tumors. Long Point Inner
Bay was selected as the least-impacted Lake Erie control site.
The Department works with a Niagara River Remedial Advisory Committee in
implementing the Remedial Action Plan. The Committee approved a delisting criterion
for the Fish Tumors BUI that reflects the Presque Isle Bay approach:
Appendix B – Public Meeting Notes B-5
“The prevalence of neoplastic liver tumors in brown bullhead is statistically
equivalent to or lower than the prevalence at Long Point Inner Bay (as
confirmed by histopathology).”
USFWS Niagara River Study
The USFWS began a study of tumor prevalence in U.S. Niagara River AOC brown
bullhead in June 2011. For purposes of sampling, the river was divided into upper,
middle and lower zones with a planned sample size of 50 fish from each zone.
The upper zone was separated into two sites, the developed shoreline (Buffalo Outer
Harbor and Black Rock Canal) and undeveloped shoreline (Strawberry Island and East
River Marsh). The middle zone was also divided into two sites, an area near the mouth
of Gun Creek at the northeast end of Grand Island and an area around Cayuga Island
that is in the vicinity of the Love Canal and 102nd Street Landfill hazardous waste sites.
The lower zone included the portion of the river between Lewiston and Fort Niagara.
Shoreline along this approximately 6.7 mile reach is not heavily developed. Although
samples were collected in four distinct areas of the reach, the fish were not kept
separate for analysis because the habitat and the degree of shoreline development
were not remarkably different among the four areas.
U.S. Geological Survey (USGS) fishery biologist Dr. Vicki Blazer was responsible for the
examination of the fish for the presence of tumors. Teams visited the AOC to process
the brown bullhead samples. Examination of the livers took place at the USGS’s
National Fish Health Research Laboratory in West Virginia.
Five of the 150 fish collected for the study were found to have liver tumors. Three of
these fish were collected in the middle zone and two in the upper zone. The incidence
of liver tumors ranged from 0% for the lower zone sample to 7.7% for the Gun Creek
sample. The overall uncorrected incidence was 3.3%.
The statistical analysis that shows that the tumor prevalence in the Niagara River AOC
and Long Point Inner Bay are equivalent is complex. The statistician who carried out the
analysis was not available to attend the meeting, but Mr. Filipski gave a very brief
overview of the analysis and results.
A simple comparison of overall tumor prevalence percentage between the AOC and the
control site would not be an appropriate approach to the statistical analysis because it
Appendix B – Public Meeting Notes B-6
would require making assumptions about the data that are not true. The statistician
used certain modelling techniques to overcome this problem.
A model in which the probability of a fish having a tumor was dependent upon the age
and gender of the fish was the best match for the data. Based on this model, the
analysis generated estimates for the tumor prevalence for the control site and the
Niagara River. Because the probability of having a tumor varies with the age of the fish,
it was necessary to choose an age at which to calculate the estimates. The average age
of bullhead in the data set was 6.4 years, so all tumor prevalence estimates were
calculated for age 6 bullhead. Gender was the other important factor, so estimates for
both male and female bullhead were calculated. The resulting estimates of Niagara
River AOC tumor prevalence for both female fish (2.0%) and male fish (1.9%) compare
favorably with the generally accepted background prevalence of two to five percent for
fish age 3 and older.
The final step of the analysis was to apply a statistical test for equivalency that takes
into account the uncertainty in the tumor prevalence estimates. The test showed that
the prevalence in the AOC is statistically equivalent to the prevalence in the control site.
This outcome shows that the delisting criterion is satisfied. Therefore, both NYSDEC
and the Remedial Advisory Committee fully support changing the status of the BUI to
“not impaired.”
Environment Canada regularly samples water at Fort Erie and Niagara-on-the-Lake to
check for a variety of chemicals. The results, which show that levels of contaminants in
the Niagara River have fallen since the 1980s, provide supporting evidence for the BUI
removal proposal. The presentation included charts showing the changes in
concentrations of PCBs and Mirex in Niagara River water over a 20 year period.
Concentrations of both fell by about 80 percent over the period.
Following removal of the Fish Tumors or Other Deformities BUI, NYSDEC and USEPA
will continue ongoing environmental programs that will ensure that the BUI’s status
remains “not impaired.” These will include completing the remediation of hazardous
waste sites, characterizing sediment within the AOC and conducting remediation where
necessary, continuing to regulate point source discharges of industrial and municipal
wastewater and storm water, regular monitoring of water quality, and coordination of the
Remedial Advisory Committee.
The presentation concluded with a review of the remaining steps in the BUI removal
process.
Appendix B – Public Meeting Notes B-7
Discussion:
Mr. Rezabek asked whether all bullhead sampled in each zone of the AOC were
collected on the same day. Mr. Filipski responded that they were collected over a short
period of time, but not in a single day. Some fish were held in the water after being
collected. The dates in the table in the presentation were the dates that the USGS
scientists visited the Niagara River to process the collected fish.
Mr. Zelazny asked Mr. Rezabek whether members of the Niagara River Station Fishing
Club are currently seeing tumors on fish they catch. Mr. Rezabek responded that they
are seeing less than they did historically. He specifically mentioned an area near the
northern end of Buffalo Outer Harbor where people historically commonly caught
freshwater drum with large skin tumors and deformed mouths. He added that he was
unable to recall the last report of skin tumor on a freshwater drum.
Conclusion:
Meeting participants had no objections to the removal of the Fish Tumors BUI based on
the outcome of the USFWS study as detailed in NYSDEC’s draft removal proposal.
Appendix C – Responsiveness Summary C-1
Appendix C – Responsiveness Summary
NYSDEC received no substantive comments on the draft of this BUI removal proposal
made available for public comment during the 30 day period from August 12 to
September 11, 2015. Therefore, no responsiveness summary is required.
Appendix D – Statistical Analysis Report D-2
Final Niagara River AOC Tumor Statistical Analysis Report
Michael A. Rutter
12/22/2015
Introduction This is the final report of statistical analysis of the Niagara River AOC bullhead tumor data. In this analysis, I concentrate on total liver and total skin tumors. Two types of analyses were conducted for this report. The first is a Bayesian analysis using the same methodology as Rutter (2010). The second analysis looks at the same data using a chi-square test of independence, a traditional approach that assumes no other covariates, for example age, affect tumor prevalence. For each of these analyses, only bullhead with all covariates measured were used in the analysis. For the Niagara River data set, 146 fish out of the 150 fish sampled were utilized. Long Point (inner bay) located near Port Rowan, Ontario, was used as the Lake Erie reference site. This was the lone reference site used in Rutter (2010). Data from the Presque Isle Bay study (2002-2007) were combined with recently collected data. Of the 243 fish collected in the reference site, 241 fish were included in the analysis.
Bayesian Analysis
Methods The Bayesian analysis utilizes a hierarchical logistic regression model to summarize tumor prevalence on bullhead in the Niagara River and compare it to tumor prevalence observed at the reference site. A logistic regression model is used when the response variable has only two possible outcomes. When analyzing tumor data, the response variable is the presence or absence of a tumor. Logistic regression also allows for the inclusion of any number of predictor variables that may affect tumor prevalence. Variables examined in this analysis include age, length, weight, and gender of brown bullhead. An important consideration for a statistical analysis of an AOC is sampling design. Traditional statistical methods assume that fish are sampled using a simple random sample approach. That is, each fish in the AOC is equally likely to be sampled. As with many sampling designs involving fisheries, this is not the case. Only certain, pre-determined locations in an AOC are selected to be sampled, and a pre-determined number of fish are usually sampled in each location. One method to statistically account for the sampling design is to use a hierarchical modeling approach (Rutter 2010). For
Appendix D – Statistical Analysis Report D-3
the Niagara River, five different sites were sampled in three different years. From a statistical perspective, since five sites were sampled, all the bullhead are not independent. By using a hierarchical approach, the correlation among brown bullhead sampled in the same location and year is accounted for. For the reference location, all the fish were sampled from Long Point (inner bay) but over multiple years. Again, the non-independence in the sampling design is accounted for with the hierarchical model. By using a hierarchical model, AOC wide tumor prevalence can be estimated by correctly combining the samples collected at each site in the AOC. The first phase of the analysis was to determine which combination of predictor variables best describes tumor prevalence. For both liver and skin tumors, hierarchical logistic regression models were fit using all possible combinations of the four predictor variables: age, length, weight, and gender. For each model, the Deviance Information Criteria (DIC, Gelman et al. (2003)) was calculated. DIC is a measure of goodness of fit used in Bayesian hierarchical models that penalizes for the number of parameters in the model. That is, as more parameters are added to the model (age, length, and gender, for example), the DIC value will increase due to the large number of predictor variables. The model with the smallest DIC is the model out of all those examined that best fits the data. Once the best set of predictor variables was selected for each type of tumor, the tumor prevalence for the reference site and Niagara River were compared. All the sites for the Niagara River were combined over all years for one estimate of tumor prevalence. One tumor prevalence was also estimated for the reference site, combining over years. Each point estimate is based on the median of the posterior distribution of tumor prevalence and can be thought of as the representative tumor prevalence for the site. The tumor prevalences were estimated for the average value of the predictor variables, combined over both data sets. It is important to compare tumor prevalence on a standardized, representative fish across locations, as comparing tumor prevalence on age nine fish at one site to an age three fish at another site does not make sense if tumor prevalence is influenced by age. It also is important to note that this standardized tumor prevalence on a representative fish is determined by the model using data collected on all fish, not just fish of that representative age. A 95% highest posterior density interval (HPDI, the Bayesian version of a confidence interval) for each tumor prevalence was reported as well. The Niagara River and the reference site were compared by calculating the difference in tumor prevalence between the two sites. The sites are considered to have statistically equivalent tumor prevalence if the 90% HPDI is contained within a given tolerance level. This procedure is called a Two One-Sided Test (TOST) of equivalency (Rutter 2010).
Appendix D – Statistical Analysis Report D-4
The tolerance level for the Niagara River AOC was created by determining what tolerance level would be required for the AOC to be equivalent with itself, a similar approach to what was done for the the Presque Isle Bay AOC (Rutter 2010).
Results
Best predictor variables For each type of tumor, 15 different hierarchical logistic regression models were fit for all combinations of the four predictor variables. Models with one, two, three, and four predictor variables were considered. No interactions of the predictor variables were considered. The results for liver tumors are given in Table 1 and the results for skin tumors are given in Table 2. Model DIC Model DIC age, gender 96.5 age, gender 292.2 age, weight 97.9 age 293.5 age, weight, gender 98.3 age, length 296.0 age, length, gender 98.8 age, length, gender 296.7 age, length, weight, gender 99.2 age, weight, gender 299.6 age 99.4 age, length, weight, gender 299.8 age, length 100.2 age, weight 302.1 age, length, weight 100.9 age, length, weight 302.7 gender 108.2 length 303.9 length 109.8 length, gender 305.2 length, gender 110.5 length, weight 309.1 weight, gender 111.8 length, weight, gender 311.1 weight 112.3 weight 311.3 length, weight 113.9 weight, gender 313.1 length, weight, gender 114.1 gender 330.1
Table 1 – DIC values for liver tumor models Table 2 – DIC values for skin tumor models
For both skin and liver tumors, the best model of those examined was the model that included age and gender as predictor variables. It should be noted that concerns have been raised in the past about including both age and length in fisheries related models, as the two variables can be highly correlated. However, for the Niagara River data, the correlation between age and length was r = 0.57 and r = 0.45 for the Long Point (inner bay) reference site. Given these levels of correlation, there are no statistical issues including both covariates in the model at the same time.
Appendix D – Statistical Analysis Report D-5
Tumor Prevalence Estimates For both types of tumors, age was an important predictor. The average age of bullhead in the data set was 6.4 years, so all tumor prevalences were estimated for age 6 bullhead. Given that gender was the other important predictor variable, estimates for both male and female bullhead are given. Point estimates of tumor prevalence are based on the median of the posterior distribution of tumor prevalence. Location and Gender 95% HPDI Median Niagara River: Female (0.0%, 25.5%) 2.0% Niagara River: Male (0.0%, 35.2%) 1.9% Long Point: Female (0.0%, 4.4%) 0.4% Long Point: Male (0.0%, 5.1%) 0.3%
Table 3 – Liver tumor prevalence
Location and Gender 95% HPDI Median Niagara River: Female (0.5%, 65.6%) 22.5% Niagara River: Male (0.3%, 79.3%) 26.7% Long Point: Female (0.0%, 12.4%) 2.2% Long Point: Male (0.0%, 19.2%) 2.6%
Table 4 – Skin tumor prevalence
TOST of Equivilance In Rutter (2010), Two One Sided Tests (TOST) of equivalence were conducted to see if AOC tumor prevalence was equivalent to the reference site. In that approach, if the 90% HPDI for the difference in tumor prevalence was within a given tolerance, the locations were considered to be statistically equivalent. For the Niagara River AOC, gender was also an important predictor of tumor prevalence, so gender specific tolerances were calculated. Since liver tumor prevalences are the lower and estimated with more certainty, they were used to create TOST tolerance levels for each gender. For females, the Niagara River TOST tolerance is ±24%. For males, the Niagara River TOST tolerance is ±34%. The 90% TOST HPDI interval for female liver tumors is (-5.4%, 18.8%), well within the Niagara River tolerance. Liver tumors for males [(-5.5%, 25.5%)] is also within the tolerance for the Niagara River, indicating statistical equivalence. Based on the Niagara River tolerances skin tumors for females [(-2.1%, 57.4%)], and skin tumors for males [(-3.1%, 66.9%)] exceed the the tolerance, indicating those tumor prevalences are not statistically equivalent.
Appendix D – Statistical Analysis Report D-6
Traditional Statistical Approach The traditional statistical approach to tumor prevalences is to use a chi-square test of independence to see if location is independent of tumor prevalence. The disadvantage of such an approach is that data such as age and gender are ignored. In the case of liver tumors, the number of tumors is low enough that a Fisher's exact test is used to test for independence. For liver tumors, there is not enough evidence (Odds Ratio=0.721, p=.754) to suggest that liver tumor prevalence depends on location. Skin tumors are plentiful enough that a chi-square test indicates there is enough evidence (χ2=50.26, p<0.0001) to suggest that skin tumor prevalence differs depending on location, with the Niagara River skin tumor prevalence being higher.
References Gelman, Andrew, John B. Carlin, Hal S. Stern, and Donald B. Rubin. 2003. Bayesian Data Analysis, Second Edition (Chapman & Hall/CRC Texts in Statistical Science). 2nd ed. Hardcover; Chapman; Hall/CRC. Rutter, Michael A. 2010. “A Statistical Approach for Establishing Tumor Incidence Delisting Criteria in Areas of Concern: A Case Study.” Journal of Great Lakes Research 36 (4): 646–655.